Elucidating the risk of cardiopulmonary consequences of an exacerbation of COPD: results of the EXACOS-CV study in Germany

Data source and study population

The WIG2 (Scientific Institute for Health Economics and Health System Research) is an anonymised healthcare claims database with longitudinal data from approximately 4.5 million individuals from different statutory health insurance funds in Germany. The database is representative of the German population, with respect to the distribution of age and sex.21 22 Information is collected in inpatient and outpatient care settings, and includes demographics, medical diagnoses, prescriptions and corresponding dispensations, and full billing information. Patients or the public were not involved in the design, or conduct, or reporting, or dissemination plans of our research.

Patients with a COPD diagnosis (International Classification of Disease, V.10, German Version, ICD-10 J44, J44.0 and J44.1) in the inpatient or outpatient setting documented between 1 January 2014 and 31 December 2018, and aged 40+ years, were included. The first COPD diagnosis defined cohort entry. Only patients with 2 years of data availability prior to cohort entry were included. In the absence of any COPD-related code prior to cohort entry, the patient was considered to be newly diagnosed. Baseline was defined as the 24-month period preceding cohort entry.

Study follow-up began on the date of cohort entry and ended upon the first occurrence of the outcome of interest, or right censoring (leaving the database or 31 December 2019).

Exposure

The exposures of interest were moderate or severe acute exacerbations of COPD. Moderate exacerbations were those managed in the outpatient setting with a course of systemic prednisolone and prednisone (Anatomical Therapeutic Chemical classification – ATC codes available in online supplemental table S1), with a dosage of 20–60 mg of prednisolone equivalent per day, for a duration of 3–15 days. Exacerbations were considered severe when they required hospitalisation and were identified using discharge codes (J44, J44.0 or J44.1 as the main OR J44.0 as the secondary discharge code) or an emergency room visit.23–27 Two exacerbations documented within 14 days of each other were counted as a single exacerbation event. Detailed definitions are provided in the online supplemental material.

Following cohort entry, a patient could contribute unexposed and exposed time. Exposed time, when patients were considered at risk of an outcome, was defined as time starting on day 1 of a moderate or severe exacerbation onset with a maximum of 365 days. On occurrence of a subsequent exacerbation, exposed time started again at day 1. Since the instantaneous rate of the outcome was expected not to be constant during the first year following an exacerbation, exposed time periods were divided into subperiods determined a priori based on previous literature15 28: days 1–7, 8–14, 15–30, 31–180 and 181–365. Unexposed time was defined as the time prior to the first exacerbation during follow-up (or prior to censoring, in the absence thereof) and all time after the 365 days following an exacerbation (figure 1).

Request permission

Outcomes

Outcomes of interest were a hospitalisation for a severe CV event, including claims for diagnosis of acute coronary syndrome, HF decompensation, cerebral ischaemia or arrhythmias (ICD-10 codes in online supplemental table S2)—or all-cause death. Outcomes were considered individually and as two composites: (1) any type of severe CV event or all-cause death, and (2) any type of non-fatal severe CV event (with death as a censoring event). Study endpoints were time to the first outcome of interest using different categorisations.

Statistical analysis

Analyses were conducted using R software, V.3.6.2. Missing codes for a given comorbidity or healthcare event were not identifiable or imputable and assumed to represent non-existent events.

Baseline characteristics were described in the entire cohort, in the subset of newly diagnosed patients and stratified by exacerbations during follow-up (at least one, or none).

To gain insight into the temporality between exacerbations of COPD and severe CV events, a swimmer plot was used. Patient subgroups were defined based on exposure and outcome status during follow-up, according to having had 0, 1, 2 or 3 moderate or severe exacerbations, and according to having experienced the outcome of a first severe CV event, including all-cause death. The swimmer plot included a bar for each subgroup, showing the typical trajectory of a patient, from cohort entry through to the end of follow-up, via exposure to exacerbations, based on median time.

Crude incidence rates of outcomes per 100 person-years were calculated with corresponding exact Poisson 95% CIs, during unexposed and exposed periods of time following an exacerbation (all exacerbations, moderate only and severe only).

Time-dependent Cox proportional hazards models were used to estimate the association between (1) the rate of outcome (any severe CV event or death; any non-fatal severe CV event; and each type of event separately) and time following an exacerbation of any severity; (2) the rate of each composite outcome and time following a moderate, or a severe exacerbation separately; and (3) the rate of a severe CV event or death following a first, a second and a third exacerbation of COPD in newly diagnosed patients. The reference period consisted of all unexposed periods as defined above, as we hypothesised that the risk of a severe CV event or death would return to the level of risk observed prior to a first exacerbation after 365 days following exacerbation onset. Sensitivity analyses were conducted to confirm this hypothesis, using only the time prior to a first exacerbation (or censoring) as an unexposed period. All models were fitted with and without adjustment for all prespecified time-invariant and time-varying confounders (online supplemental table S3). No adjustment was made for multiple comparisons, as the outcomes were deemed interdependant.

Results

The cohort consisted of 126 795 individuals, including 46 467 (36.6%) with a new COPD diagnosis (figure 2).

Request permission

In the overall population, the average age was 66.5 years, 60% were men and 21% had a history of at least one exacerbation during the 24 months prior to cohort entry. The most common comorbidities were hypertensive diseases (75.2%), hyperlipidaemias (56.6%) and anxiety disorders (45.8%). The newly diagnosed patients were younger (mean of 64.5 years) and had fewer comorbid conditions (table 1; medication use is detailed in online supplemental table S4). Baseline characteristics of patients categorised by exacerbation status during follow-up are provided in online supplemental table S5.

Over a mean follow-up of 36.0 months (SD, 21.1), 37 701 (29.7%) patients had at least one moderate exacerbation, and 34 232 (27.0%) had at least one severe exacerbation prior to the outcome, or censoring.

48 982 (38.6%) patients experienced either a non-fatal severe CV event (28 568, 22.5%), or died without experiencing any severe CV event (20 414, 16.1%) (table 2). The non-fatal severe CV events that most commonly occurred first were decompensated HF (10 486, 21.4%) and acute coronary syndrome (7109, 14.5%).

Temporal relationship of exacerbation and severe CV events during follow-up

In patients who had a severe CV event or died during follow-up, the time since cohort entry in those with, or without an exacerbation prior to the CV event or death, varied. In patients with no exacerbation, the median time to the first CV event or death was 10.5 months; in patients who experienced at least one exacerbation prior to having a CV event or death, the median time elapsed between an exacerbation and the event was 1 month following a first exacerbation, 0.7 months following a second exacerbation and 0.6 months following a third exacerbation (figure 3).

Request permission

Rate of a first severe CV event following the onset of an exacerbation (any severity)

Crude incidence rates of a first severe CV event or death following an exacerbation of any severity, a moderate or a severe exacerbation, are reported in online supplemental table S6.

Figure 4 displays the adjusted rate of a first severe CV event associated with time periods following the onset of an exacerbation of any severity. The rate of a severe CV event or death was increased during the entire year following an exacerbation, with the highest rate observed during the 1–7 day period (adjusted HR, aHR 8.6, 95% CI 8.3 to 8.9). An increased rate was also found for non-fatal severe CV events, but of lower magnitudes (1–7 day period aHR 5.2, 95% CI 4.9 to 5.5). The rate increase persisted during the entire year of exposed time.

Request permission

In sensitivity analyses excluding time beyond 365 days following an exacerbation from the unexposed reference period, the estimated HRs were very similar (results not shown). As a result, all models used as reference periods the combination of time prior to a first exacerbation during follow-up (or censoring, in the absence thereof) and all time exceeding 365 days following an exacerbation.

When examining the rate of each type of severe CV outcome individually in the time periods following any exacerbation onset (online supplemental figure S1), there was an increased rate of all outcomes, however with different durations of elevation in risk. The rate of all-cause death was significantly elevated over the entire year following an exacerbation (aHR 1.60 (95% CI 1.51 to 170) during a 181–365 day period); the rates of HF decompensation and of cerebral ischaemia were significantly increased for up to 180 days, while the increase in the rates of the acute coronary syndrome was of shorter duration (≤30 days).

Risk of a first severe CV event following the onset of a moderate or a severe exacerbation

The rate of a first severe CV event or death was increased following the onset of a moderate exacerbation and remained elevated for up to 6 months (figure 5) with aHR of 1.17 (95% CI 1.05 to 1.31) in the 1–7 day period, and an aHR of 1.08 (95% CI 1.04 to 1.13) in the 31–180 day period. The rate of a first non-fatal severe CV event following a moderate exacerbation followed a similar pattern.

Request permission

The rate of a first severe CV event or death following the onset of a severe exacerbation was increased to a greater extent (compared with periods of time following a moderate exacerbation), and the increased rate persisted for the entire year. In analyses of the composite outcome including death, the latter event contributed substantially to the overall increase in risk, namely in the immediate period (1–7 days) following the onset of a severe exacerbation: aHR was 15.84 (95% CI 15.26 to 16.45) when including deaths, and aHR was 8.51 (95% CI 8.03 to 9.01) when death was a censoring event.

Risk of a first severe CV event following the onset of a first, second and third exacerbation

During follow-up, of the 46 467 patients considered as newly diagnosed, 27 558 (59.3%) had no exacerbation and 18 909 (40.7%) had at least one exacerbation. In addition, 7974 (17.2%) patients had at least two exacerbations and 4428 (9.5%) had at least three exacerbations. Of the 18 909 patients with at least one exacerbation, 4843 (25.6%) experienced at least one severe CV event or died, and 2512 (13.3%) patients experienced a non-fatal severe CV event during the 365 days following the onset of an exacerbation (exposure).

Figure 6 displays the aHR for the rates of a severe CV event associated with periods following a first, a second and a third exacerbation (of any severity). Following the first exacerbation, an increase in risk of the outcome was observed. Higher increases in rates were observed in periods of time following a second exacerbation, compared with time periods following a first exacerbation. The outcome rate increase following a second exacerbation was of similar magnitude compared with periods following a third one.

Request permission

Key findings

The present study explored the risks of severe CV events experienced by people with COPD in time periods following an exacerbation, using a large population-based cohort of more than 126 000 people living with COPD in Germany. We found that rates of a range of severe CV events or death are increased for at least a year following exacerbation onset, for severe exacerbations and up to 180 days following a moderate one. The rate increase was present for prevalent as well as incident, newly diagnosed patients.

One important finding of our study is the high burden of severe CV disease and events in this population. The prevalence of cardiac conditions was substantial, with, for example, 37% of patients with a history of ischaemic heart disease, or 28% with a history of HF. In a previous study conducted on a cohort retrieved from the same database,10 the prevalence of cardiac diseases was possibly underestimated (22% of ischaemic heart disease and 13% of HF) because only the 12-month period prior to cohort entry was used to identify comorbid conditions. In the present study, the entire available lookback period was used, thus providing more sensitive information on the burden of CV diseases in patients living with COPD. Therefore, the proportions of patients with comorbidities appear relatively high in our study but nevertheless align with reported ranges of other international or national studies.29 30 This consistency of comorbidity rates between other studies and our observations also confirms the representativeness of our population. Moreover, during an average follow-up of 3 years, one-fourth of the cohort had at least one hospitalisation for a severe non-fatal CV event and 16% of patients died.

Another important finding is the association of moderate exacerbation with an increased risk of a severe CV event. This is an important result given the frequency of moderate exacerbations. In the present study, nearly one-third of patients had at least one moderate exacerbation prior to having a CV event, or censoring. However, the risk increase was found to be more prolonged than previously suggested,14 15 28 as it remained elevated for up to 6 months. The shorter duration of elevated risk identified in the post-hoc analyses of the IMPACT trial28 31 may be explained by the close monitoring of patients included in trials, potentially minimising the risk of adverse events, differences in the study populations (eg, underlying cardiac comorbidity) and trial duration. From a pathophysiological perspective, persistent post-exacerbation inflammation may contribute to elevated systemic inflammation and non-recovery of symptoms following the acute phase of the exacerbation.32 In this regard, our findings support the concept of an immediate period of higher vulnerability after the onset of an exacerbation, as reported by Hurst and colleagues.33

To our knowledge, the EXACOS-CV programme is the first to explore the impact of exacerbations in patients with a new diagnosis of COPD in a time-dependent manner. Following a first recorded exacerbation after a new diagnosis of COPD, we observed a sevenfold increased risk of severe CV event or death during the acute phase (1–7 days post onset). The increase in risk persisted over the entire year following a first exacerbation highlighting the burden of exacerbations even at an early stage of disease management. COPD diagnosis is usually made years after its onset and thus, the so-called first exacerbation is not necessarily the very first one. Indeed, 17% of the newly diagnosed patients in our cohort experienced at least one possible exacerbation prior to cohort entry. However, the period of COPD diagnosis is important from a clinical standpoint because a thorough examination of the patient prognosis factors is recommended by national and international professional societies.2 Our results confirm the need for an assessment of CV risk factors as part of the initial assessment of disease and quantification of future risk, proposed in an ongoing Quality Improvement Initiative.34

In this population, we also explored the risk of a severe CV event or death following each subsequent exacerbation of COPD. Our results show that the median time elapsed between an exacerbation and a first severe CV event gets shorter following each subsequent exacerbation (1 month, 0.7 months, 0.6 months in patients with 1, 2 and 3 exacerbations during follow-up, respectively). In terms of risk, we observed an 11-fold increase in the 1–7 days following a second exacerbation (vs a 7-fold increase following a first one), suggesting an increased susceptibility to having a severe CV event following a second exacerbation. However, no difference was found between the risk following a second, and the risk following a third exacerbation. It is possible that a cohort effect led to underestimating the risk associated with time following a third exacerbation, as the patients most likely to have a severe CV event would tend to have it earlier and leave the cohort prior to experiencing a third exacerbation. Despite adjustment on time-varying confounders, the different underlying risk profiles of patients having a severe CV event earlier versus later could not be totally accounted for.

Previous studies exploring the risk of a CV event following an exacerbation were conducted on patients with severe COPD selected for clinical trials,11 16 28 or having the outcome of interest.12 14 15 18 The present study included a heterogeneous unselected cohort of patients with COPD irrespective of their exacerbation or CV status, which represents clinical care. Moreover, the inclusion of patients with a new diagnosis of COPD allows for further characterisation of patients in the course of the disease and makes our results more generalisable to the German population of patients with COPD.

Strength and limitations

An important limitation was the inadequate measurement of behavioural risk factors. As is very frequently the case in claims databases, obesity and smoking status are under-reported as they do not lead to reimbursement. In our study, only 40% of patients were observed to have a history of smoking. Other prospective observational cohort studies in Germany report a proportion of 92%35 and 79%36 of patients with a smoking history. The measurement error incurred by missing data may have led to residual confounding.

Further, as a common limitation of observational retrospective cohort studies using routinely collected health data, there is a risk of misclassification of endpoints also in our study. This refers to (1) misclassifications of exacerbations that are not present; (2) undetected or under-reported mild-to-moderate exacerbations; (3) misclassification of elective hospital stays as acute events; (4) acute events happening during hospital stays with other main diagnosis; (5) inverse temporality, when an exacerbation leads to hospitalisation, but severe CV event is chosen as the main diagnosis.

Clinical implication

Our results emphasise the high burden of cardiopulmonary risk, based on respiratory events of exacerbations and severe CV events, in patients living with COPD. The observed elevation in risk of adverse severe CV outcomes following a moderate exacerbation, or even following a first recorded exacerbation in patients with newly diagnosed COPD clearly illustrates that the prevention of such events should be a priority in COPD management. The substantial increase in risk in the immediate period following the onset of an exacerbation highlights the urgency to provide proactive care to people experiencing an acute respiratory event. Moreover, the persistence of the increased risk over time indicates the need for a prolonged and multidisciplinary monitoring of patients in the year following the onset of an exacerbation.